Title:

Reconciling Analytics-based Biomedical Research and Individuals' Privacy: A Differential Privacy Based Approach

Abstract:

Large-scale mining of electronic medical records might clarify what treatment approaches actually work (evidence-based medicine) and clarify what factors foretell or contribute to disease, potentially allowing us to fend off disease before it even starts. The routine inclusion of genetic information offers the possibility of even deeper insights and personalized therapies.

On the other hand, even simple, routine analytics have the potential to compromise individuals' privacy, to a greater degree than is generally realized. Yet guarding privacy by conducting all research behind a Chinese Wall of IRB-imposed safeguards will significantly slow down biomedical progress, at significant cost to society.

Society will eventually need to debate these issues and decide how to balance them. As a first step, the issues and potential safeguards must be clarified. In this talk, I'll explain some realistic risks to individuals' privacy from routine publication of results from biomedical studies. I'll describe how differential privacy can help with this particular problem, present other examples of where differential privacy can be useful in biomedical research, and describe where it falls short as a protection technique in practice.

Speaker:

Dr. Marianne Winslett,
Department of Computer Science,
University of Illinois at Urbana-Champaign